Interpretation biases in social anxiety: Response generation, response selection, and self-appraisals

ARTICLE IN PRESS

Behaviour Research and Therapy 45 (2007) 1505–1515 www.elsevier.com/locate/brat

Interpretation biases in social anxiety: Response generation, response selection, and self-appraisals Jonathan D. Hupperta,, Radhika V. Pasupuletia, Edna B. Foaa, Andrew Mathewsb a

Center for the Treatment and Study of Anxiety, University of Pennsylvania School of Medicine, 3535 Market St, Suite 600N, Philadelphia, PA 19104, USA b Kings College, London, UK Received 16 May 2006; received in revised form 5 December 2006; accepted 23 January 2007

Abstract Cognitive theories propose that the resolution of ambiguity is related to the maintenance of social anxiety. A sentence completion task was used to examine how individuals high (n ¼ 26) and low (n ¼ 23) in social anxiety resolve ambiguous social sentences. Individuals were asked to generate as many responses as came to mind for each sentence, and then to endorse the response that best completes the sentence. Total responses, first responses, and endorsed responses were examined separately. Results indicated that high anxious individuals had more negative and anxious responses and fewer positive and neutral responses than low anxious individuals on all sentence completion measures. In contrast, a self-report measure of interpretation bias indicated that more of negative and anxious appraisals were related to social anxiety, while positive and neutral appraisals were not. Results are discussed in terms of a multi-stage processing model of interpretation biases. r 2007 Elsevier Ltd. All rights reserved. Keywords: Cognitive mechanisms; Interpretation bias; Social anxiety; Depression

Introduction Cognitive-behavioral theories of social phobia propose that the manner in which one interprets an ambiguous social interaction plays an important role in the maintenance of social anxiety (Clark, 2001; Huppert & Foa, 2004; Rapee & Heimberg, 1997). Given that the feedback in social interactions is usually ambiguous (i.e., one rarely receives uncensored praise or criticism), the resolution of this ambiguity provides an important source of information to the individual in how one construes oneself as a social being. If one ultimately interprets ambiguous social feedback as positive, then he/she is more likely to see himself/herself as more competent whereas if the interpretation is ultimately negative, he/she is more likely to see himself /herself as incompetent and will therefore be more anxious. However, if the resolution of ambiguity involves the elicitation of multiple interpretations in response to an ambiguous scenario followed by the selection of a Corresponding author. Tel.: +1 215 746 3327; fax: +1 215 746 3311.

E-mail address: [email protected] (J.D. Huppert). 0005-7967/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.brat.2007.01.006

ARTICLE IN PRESS 1506

J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

single interpretation, then it is possible that both positive and negative responses will be generated initially. One question that arises from this reasoning is whether individuals with social anxiety differentially generate responses to ambiguous scenarios or if they only differentially select negative or positive interpretations. No research on interpretation biases in social anxiety has examined generation of multiple interpretations of a single ambiguous social scenario. Instead, single interpretations are presented within reaction time paradigms (Amir, Beard, & Przeworski, 2005; Hirsch & Mathews, 2000), single responses to open-ended scenarios are examined with the assumption that the first response to come to mind is equivalent to the endorsed response (Franklin, Huppert, Langner, Leiberg & Foa, 2005; Stopa & Clark, 2000), or multiple interpretations are passively presented and rank ordered (Amir, Foa, & Coles, 1998; Stopa & Clark, 2000) or rated (Huppert, Foa, Furr, Filip, & Mathews, 2003). Thus, the main goal of this study is to examine whether multiple interpretations of ambiguous social scenarios evoke a lack of positive interpretations, the presence of negative interpretations, or both, and how these responses relate to first and explicitly endorsed responses. In one of the first studies to examine open-ended responses to ambiguous social scenarios, Stopa and Clark (2000) investigated interpretation bias associated with social anxiety by presenting participants with ambiguous social and ambiguous non-social scenarios and eliciting a single written interpretation for each scenario. Participants with social anxiety generated more negative interpretations of ambiguous social scenarios than participants with other anxiety disorders and controls. The findings from Stopa and Clark were replicated by Franklin et al. (2005), who showed that untreated patients exhibited a bias towards negative interpretations while successfully treated socially phobic patients did not exhibit a negative interpretation bias. In contrast, some studies have reported that individuals with social anxiety lack a positive interpretation bias, but do not evidence a negative interpretation bias (Amir, Beard, & Bower, 2005; Hirsch & Mathews, 1997, 2000). In the earliest of these studies in socially anxious individuals, Hirsch and Mathews (1997) found evidence for the role of lack of positive bias in maintaining social anxiety. A reaction time paradigm was utilized to investigate immediate responses to positive or negative interpretations of ambiguous scenarios. Individuals low in social anxiety responded faster to the positive words than individuals who were high in social anxiety, but the two groups did not differ in their speed of responses to negative words. Hirsch and Mathews (2000) replicated these results with clinically socially anxious participants. These findings were interpreted as indicating that socially anxious individuals lack positive on-line inferences to ambiguous social scenarios. Similar findings have been reported by others utilizing other reaction time paradigms (e.g., Amir, Beard, & Przeworski, 2005). To examine these seemingly conflicting results, Huppert et al. (2003) investigated both the presence of a negative bias and the lack of a positive bias separately by asking individuals to rate the similarity of positive and negative interpretations of previously presented ambiguous scenarios as a recall task. The results showed that: (1) severity of social anxiety was related to interpreting an ambiguous social situation negatively; (2) lack of positive bias was more related to general anxiety than to social anxiety; and (3) negative and positive interpretation biases were not strongly related. The authors suggested that these two biases should be measured separately, and that both may play a role in the maintenance of social anxiety. To our knowledge, no studies have utilized sentence completion tasks to examine interpretation processes in social anxiety, although such tasks may provide an adequate method for elucidating such biases. Indeed, sentence completions tasks are frequently used to tap into multiple semantic networks that are activated by incoming stimuli (Kutas & Hillyard, 1984; Williams & Colombo, 1995). In fact, there is evidence that several meanings are activated when a participant is exposed to an ambiguous word (Gernsbacher, 1991), and the same may be true for ambiguous sentences. Once multiple meanings are activated, processes of enhancement and/or suppression lead to a final resolution of meaning (c.f., Gernsbacher & St. John, 2001; MacLeod, Dodd, Sheard, Wilson, & Bibi, 2003; McNamara & McDaniel, 2004 for a detailed discussion regarding the controversies regarding the concept of suppression). Regardless of the mechanism involved, it appears that multiple semantic networks are activated when presented with incoming ambiguous stimuli and that the individual ultimately selects one meaning over another. Therefore, a sentence completion paradigm in which multiple responses to each sentence are obtained may be utilized to stimulate the multiple networks. The present study examined interpretation biases in social anxiety utilizing a sentence completion task and a self-report measure. This procedure extends the work of Stopa and Clark (2000) and Franklin et al. (2005), where participants were instructed to write only the first response that came to mind. First, participants were

ARTICLE IN PRESS J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

1507

presented with ambiguous sentences in which the last word of the sentence was omitted. In order to investigate multiple associations that may be activated in an ambiguous situation, participants were instructed to come up with as many one-word answers in the order that they come to mind. In order to examine response selection, participants were also asked to endorse the response that they thought would best complete the sentence. By including multiple responses to each sentence and examining the first response, all responses, and the endorsed response, we are able to determine whether (1) the first response reflects similar biases as endorsed responses, which utilize more elaboration and reflection, (2) whether individuals are likely to come up with both positive and negative responses to a given ambiguous stimulus, and (3) whether the first, overall, and endorsed responses converge in their reflecting either presence of negative or lack of positive bias. Finally, we believed that the endorsed response, while entailing self-reflective processing, was still less subject to social desirability (i.e., attempts to present oneself in a positive light) than a self-report measure, and would be more likely to evidence a lack of positive and the presence of a negative bias. All the responses were coded by independent raters. In addition, a subset of the sentences was utilized to create a self-report measure of interpretation bias. Socially ambiguous sentences were presented twice, with the last word resolving the sentence in either a positive or negative way. Participants rated how much each statement applied to them. The last words of the sentences were intended to be self-resolutions to ambiguity, as it is unlikely that individuals typically receive unambiguous feedback in such circumstances. Thus, the more likely a resolution was rated as applying to them, the more biased they were towards that resolution. We therefore viewed this as a self-report measure of interpretation biases. It was predicted that socially anxious participants would generate and select more negative or anxiety responses and fewer positive or neutral responses than participants who were not socially anxious. We predicted a similar relationship on the self-report measure. The relationship among methods of understanding resolution of ambiguity was examined, with the idea that the total responses may be a better representation of the response generation stage of disambiguation while endorsed and first responses will be more representative of response selection. The self-report measure of bias was expected to be related to response generation and selection. Finally, given accumulating interest regarding the specificity of interpretation bias regarding social anxiety in comparison to depression (c.f., Gotlib et al., 2004; Gilboa-Schechtman, Presburger, Marom, & Hermesh, 2005), we examined the impact of depression on group differences in the bias ratings (i.e., different coded responses and the self-report measure of interpretation bias). Methods Participants Participants were recruited through newspaper advertisements, newsgroup postings, and word of mouth. Advertisements stated that the individual would be paid $40 to participate for a 2-h experiment on thoughts and emotions. No other information was given. A research assistant screened individuals who contacted the Center for the Treatment and Study of Anxiety (CTSA) by phone or e-mail expressing interest in participating. Inclusion and exclusion criteria were determined by the phone screen interview: knowledge of how to use a computer, no reported substance abuse, bipolar disorder, psychosis, or suicidality, and a Social Phobia Inventory (SPIN; Conner et al., 2000) score below 10 or above 20.1 These cutscores were determined by ROC analyses reported in Conner et al. (2000). Of 155 individuals screened, 59 met criteria and participated in the study. Of the individuals who did not participate, reasons for exclusion included not meeting high or low SPIN criteria (n ¼ 33), diagnosis of schizophrenia or bipolar disorder (n ¼ 21), being unable to use a computer (n ¼ 17), or other/unknown reasons (n ¼ 25). Measures Sentence Completion: Eighty sentences related to ambiguous social scenarios were developed by the research team. The sentences were then circulated to experts in social anxiety who provided feedback and revised 1

After recruiting approximately 35 participants, we raised the SPIN cutoff to 30 to decrease chances of false positives for social anxiety.

ARTICLE IN PRESS 1508

J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

according to the experts’ comments. All sentences described experiences within social situations and remained ambiguous until the last word. For the purposes of the sentence completion task, the last word was eliminated from the sentence (e.g., ‘‘As you walk to the podium, you notice your heart racing, which means you are _____’’ or ‘‘As you give a speech, you see a person in the crowd smiling, which means that your speech is _____’’). The final set of sentences was then recorded into .wav files by the first author. Instructions were provided to participants as follows: (1) Click the ‘‘Start’’ button once to listen to the first sentence. For other sentences, just click the ‘‘Next Sentence’’ button. (2) Complete each sentence with as many single words (not phrases) that come to mind in the order that you think of them. (3) Place an asterisk next to the word that you feel best completes the sentence. (4) When you are finished, click the ‘‘Next Sentence’’ button and continue. Individuals would then type in each word in a box on the computer screen. Once they pressed the ‘‘Next Sentence’’ button, the words were recorded into a database and the next sentence was played. Sentences were presented in random order. All words were then coded by one of two research assistants, blind to participant status, into one of five possible codes: positive, neutral, negative, anxious, non-codable/ambiguous. Coding was made separately for each word in each sentence, as a word could be neutral in one context and negative in another. Guidelines for each of the codings were as follows: positive coding suggested a good attribute (excited or eager for the first example above; interesting, funny or agreeable for the second example); neutral was coded for words that did not reveal a strongly positive or negative attribute, but completed the sentence accurately (active or hot for the first sentence; OK or tolerable for the second sentence); negative coding suggested undesirable attributes (confused or overwhelmed for the first sentence; ridiculous or stupid for second sentence); anxious coding was for words specifically related to anxiety (nervous, anxious, scared, afraid). This category was created following the guidelines of Stopa and Clark (2000), and was separate from negative codes in order to distinguish negative evaluations from descriptions of symptoms or anxiety per se. Each response was coded with only one category. Overall, 16,969 responses were recorded and coded. Both coders coded a subset of 632 responses for training, and then coded 739 responses for reliability determination. After training, coders were able to obtain 80% agreement on codes. All words which coders thought difficult to code were categorized by group consensus of the two coders and the first author. Scores for the total words (hereafter total) were adjusted for the total number of responses by using the residual of a regression predicting the number of responses for each category on the total number of responses. This was done in order to allow each of the participants to equally contribute to group means. Otherwise, individuals with fewer responses would contribute to mean response scores less than individuals with significantly more responses. By taking residuals instead of percentages, category scores remained independent while adjusting for number of responses. First and endorsed words did not require the same adjustment, given that a single response was used from each sentence for these variables. Self-Report Measure of Interpretation Bias (SMIB2): The SMIB was derived from 23 of the 80 sentences described in the sentence completion task. These sentences were selected as representing all of the different topics or scenarios reflected in the larger set of sentences. For each of the 23 ambiguous sentences, endings were added to the sentence stems to create a 46-item questionnaire (using each stem twice), which asked individuals to imagine themselves in the scene described by each sentence and to rate themselves on a 0–4 scale in terms of ‘‘how likely each sentence would apply to you’’. Scores of 0 were equal to ‘‘not at all likely to apply to me’’ and 4 were ‘‘definitely likely to apply to me’’, with intermediate anchors for 1–3. In order to create subscales that matched the sentence completion coding, subscales were derived by having the raters code the sentences that were used in the self-report measure. Of the 46 sentences, 7 were coded as positive, 17 as neutral, 3 as anxious, and 19 as negative. In order to make the scales comparable, we used the mean score for each of 2

SMIB available from the first author.

ARTICLE IN PRESS J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

1509

the four subscales. Alphas were moderate to high for each of the scales (positive ¼ .69, neutral ¼ .72, anxiety ¼ .67, negative ¼ .93).3 Liebowitz Social Anxiety Scale—Self Report (LSAS; Liebowitz, 1987): The LSAS is a measure of social anxiety, which asks individuals to rate both fear and anxiety, which has been shown to have good psychometric properties (Baker, Heinrichs, Kim, & Hofmann, 2002; Fresco et al., 2001). ROC analyses of the LSAS suggest cutscores of 430 for social anxiety disorder, and 460 for generalized social anxiety disorder (Mennin et al., 2002). Social Interaction Anxiety Scale (SIAS; Mattick & Clarke, 1998): The SIAS is a commonly used measure of social anxiety, which evaluates the severity of social anxiety in interpersonal situations and has been shown to have good psychometric properties in multiple samples (Mattick & Clarke, 1998; Ries et al., 1998; Safren, Turk, & Heimberg, 1998). Social Phobia Inventory (SPIN; Conner et al., 2000): This is a 17-item measure of social anxiety that asks about a range of social interactions, fears of embarrassment, and discomfort with physical symptoms of social anxiety. It has been used in clinical and non-clinical samples, and the psychometrics have been found to be sound (Conner et al., 2000), and was used as the screening instrument here. Depression, Anxiety, and Stress Scales—21-item version (DASS; Lovibond & Lovibond, 1995): The DASS is comprised of three subscale developed to evaluate anxiety, depression, and stress, as described by the tripartite model of affect (Watson et al., 1995). Its psychometric properties have been shown to be good in clinical (Antony Bieling, Cox, Enns, & Swinson, 1998; Brown, Chorpita, Korotitsch, & Barlow, 1997) and non-clinical (Lovibond & Lovibond, 1995; Crawford & Henry, 2003) populations. Marlow-Crowne Social Desirability Scale—short form version 2 (SDS; Strahan & Gerbasi, 1972). The Marlow-Crowne Social Desirability Scale is a commonly used measure of social desirability. The short form (X2), which has been shown to correlate highly with the long form and has psychometric properties similar to other short forms (Fischer & Fick, 1993) was used for this study. Procedures Upon arrival at the CTSA all participants signed consent to participate in the study after reading the consent form. The order of presentation of tasks was fixed in order to attempt to minimize biasing responses on the sentence completion and grammatical decision tasks. The order of tasks was as follows: sentence completion, DASS, SDS, a grammatical decision task (used as a pilot for a future experiment and not discussed here), LSAS, SIAS, and the SMIB. Measures were completed and responses recorded via a Dell PC computer using Microsoft Access. Data analyses Data were examined using separate MANOVAs for each subscale (positive, negative, anxious, neutral) with the three ratings from the sentence completion task (total, first, endorsed) as the dependent variables and group (high vs. low socially anxious) as the independent variable and depression as the covariate when applicable. As we had no hypotheses regarding the ambiguous/other response category, we did not examine any differences between groups on this category of responses. Signficant results were followed with t-tests or ANCOVAs. In addition, we examined correlations among sentence completion ratings, and the SMIB. Alpha was set to po.01 for all analyses. Results Participant analyses Of the 59 participants who signed consent for the experiment, 10 were excluded from data analysis for the following reasons (n’s in parentheses): unable to use a computer (1); screened high, but self-reports suggested 3

The SMIB was originally designed to measure positive vs. negative bias, but after development of the coding system for the sentence completion task, it was divided into the same categories to allow for parallel analyses.

ARTICLE IN PRESS J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

1510

low social anxiety (3); screened low, but self-reports suggested high (1); contacted for a screen two times, and was screened once high, once low (this was determined only after participation) (1); SDS score 47 (4). There were no differences between included and excluded individuals in group status (high vs. low social anxiety), sex, and all self-report measures except SDS (all other p’s4.05). Data from the remaining 49 participants were used for all further data analysis. Of the 49 participants, 26 were in the high socially anxious group (hereafter HSA), and 23 were in the low socially anxious group (hereafter LSA). The LSA group did not differ from the HSA group on any demographics (all p’s4.20) including age (29.3, SD ¼ 15.5 vs. 34.4, SD ¼ 12.6), sex (65% vs. 58% female), ethnicity (35% vs. 46% Caucasian, 30% vs. 27% African American, 17% vs. 15% Asian, 18% vs. 12% Other), marital status (70% vs. 69% single) or employment status (43% vs. 42% students, 22% vs. 11% working full time, 13% vs. 15% working part time, and 22% vs. 31% unemployed). MANOVA’s and follow-up tests showed that the LSA group had significantly lower scores on the SPIN, LSAS, SIAS, and DASS scores than the HSA group (F (8, 39) ¼ 27.9, po.001; see Table 1 for pairwise comparisons). There was no group difference on the SDS (p4.20). Sentence completion task Overall descriptives of total, total adjusted, first, and endorsed responses are presented in Table 2. Separate MANOVAs and follow-up tests conducted for positive, negative, anxious, and neutral responses showed group differences (F negative (3, 45) ¼ 7.7, po.001, partial Z2 ¼ .34; F positive (3, 45) ¼ 10.4, po.001, partial Z2 ¼ .41; F anxious (3, 45) ¼ 3.3, po.05, partial Z2 ¼ .18; F neutral (3, 45) ¼ 6.0, po.001, partial Z2 ¼ .29; see Table 2 for pairwise comparisons). As expected, individuals in the LSA group gave more positive and neutral responses and fewer negative or anxious responses on total, first and endorsed responses than the HSA group. Within subjects tests revealed that within the HSA group, there were more negative responses than neutral responses and more neutral responses than positive responses (all t’s43.4, p’s o.01). Within the LSA group, there were no differences in total number of negative, neutral, or positive responses. Nor were there differences between positive and negative first or endorsed responses (t’so2, p’s4.05), though there were more neutral than positive or negative first and endorsed responses (t’s43.5, p’so.01, except first neutral vs. first negative t (22) ¼ 2.7, po.05). Both HSA and LSA groups had fewer anxious responses than negative, neutral, or positive responses (t’s43.9). Depression was added in as a covariate in each of the between groups MANOVAs in order to determine whether any of the responses were uniquely related to social anxiety. MANOVAs for groups differences on positive responses remained (F positive (3, 44) ¼ 2.8, p ¼ .051, partial Z2 ¼ .16) while negative, anxious, and neutral responses all became non-significant (F’so1.1, p’s4.40). Follow-up analyses for the positive responses Table 1 Scores on symptom measures for high and low socially anxious groups Low anxious (n ¼ 23)

F ((1, 48))

High anxious (n ¼ 26)

Mean

SD

Mean

SD

5.4 24.1 15.3

3.4 12.2 8.7

39.7 83.5 48.7

11.1 28.8 16.2

200.4** 84.3** 77.4**

DASS-21 Depression Anxiety Stress

3.2 1.9 5.9

1.9 2.1 2.9

10.4 7.8 13.2

5.7 5.0 4.2

33.6** 28.3** 48.8**

SDS

4.0

1.9

4.1

1.9

.05

SPIN LSAS SIAS

SPIN—Social Phobia Inventory (administered for screening purposes); LSAS—Liebowitz Social Anxiety Scale; SIAS—Social Interaction Anxiety Scale; DASS-21—Depression, Anxiety, and Stress Scales; SDS—Social Desirability Scale. *po.01, **po.001.

ARTICLE IN PRESS J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

1511

Table 2 Sentence completion task results for high and low socially anxious participants Sentence completion

Negative words Unadjusted totala Total First Endorsed Positive words Unadjusted totala Total First Endorsed Anxious words Unadjusted totala Total First Endorsed Neutral words Unadjusted totala Total First Endorsed

Low anxious (n ¼ 23)

High anxious (n ¼ 26)

Mean

SD

Mean

SD

76.4

28.4

102.3

42.1

12.4 18.7 17.6

19.2 6.6 5.3

11.0 28.2 28.5

33.0 8.3 9.7

64.8

20.9

40.6

20.9

13.5 18.8 20.0

15.8 4.9 4.6

11.9 12.8 12.6

17.1 5.0 5.4

15.8

7.1

20.9

9.2

2.5 5.3 5.0

6.6 2.7 2.3

2.2 7.7 7.1

7.6 3.4 2.6

71.4

26.2

52.2

26.0

11.1 25.4 25.2

15.4 6.0 5.2

9.9 18.5 17.9

19.5 6.9 6.3

F (1, 48)

Effect size (Z2)

8.8* 19.2** 23.4**

.16 .29 .33

28.9** 17.8** 26.6*

.38 .28 .36

5.3! 7.4* 9.1*

.10 .14 .16

17.3** 14.1** 19.1**

.27 .23 .29

!

po.05, *po.01, **po.001. a Unadjusted total—all words per participant coded (not included in analyses); Total—total words in category adjusted for total words given by each participant; First—first words coded, Endorsed—endorsed words coded.

showed that differences between high and low subjects were significant for total adjusted (F (1, 46) ¼ 8.1, po.01; Z2 ¼ .15), first (F (1, 46) ¼ 5.1, po.05; Z2 ¼ .10), and endorsed responses (F (1, 46) ¼ 5.8, po.05; Z2 ¼ .11). The two groups did not differ in the average number of answers given per sentence (mean LSA ¼ 3.7, SD ¼ 1.2; mean HSA ¼ 3.9, SD ¼ 1.6; t (47) ¼ .4, p ¼ .72), in the average location of the endorsed word within the subject-generated list (mean LSA ¼ 2.0, SD ¼ .5; mean HSA ¼ 2.0, SD ¼ .7; t (47) ¼ .1, p ¼ .89), or in the percent of first words that were endorsed (mean LSA ¼ 52.5, SD ¼ 20.9; mean HSA ¼ 55.6, SD ¼ 23.9; t (47) ¼ .48, p ¼ .63). Self-report results A 4 (sentence type) by 2 (group) ANOVA revealed a group by sentence interaction (F (3, 45) ¼ 12.4, po.001). This interaction was examined in two ways. The anxious group scored higher than the non-anxious group on the negative and anxious subscales of the SMIB, and there was a trend for them to score lower on the positive subscale (see Table 3). There were no group differences on the neutral subscale. Within subject analyses revealed that there were no differences among scales in the high group (t’s (25)o1.6, p’s4.10), and that positive and neutral scores were significantly higher than anxious and negative scores in the low group (t’s (22)46.0, p’so.001).When covarying out depression, the difference between groups for negative interpretations (F (1, 46) ¼ 8.1, p o. 01, Z2 ¼ .15) and anxious interpretations (F (1, 46) ¼ 8.6, po.01; Z2 ¼ .16) remained significant. However, the trend in the difference between groups on the positive subscale

ARTICLE IN PRESS J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

1512

Table 3 Mean item ratings on the SMIB for high and low socially anxious participants SMIB

Low anxious (n ¼ 23)

High anxious (n ¼ 26)

F (1, 48)

Effect size (Z2)

Negative Anxious Positive Neutral

1.1 1.2 2.5 2.2

2.1 2.4 2.2 2.3

38.2** 36.3** 4.1! 1.0

.45 .44 .08 .02

0.4 0.7 0.6 0.5

0.7 0.7 0.6 0.4

! po.05, *po.01, **po.001. SMIB—Self-Report Measure of Interpretation Bias.

became non-significant. Depression also had a significant relationship with negative (F (1, 46) ¼ 15.8, po.01; Z2 ¼ .26) and anxious interpretations (F (1, 46) ¼ 10.9, p o. 01, Z2 ¼ .19), but not with positive (p4.15, Z2o.04). Correlational analyses The correlations among total, first, and endorsed words were high within each category (all p’s o.01; negative r’s (49) ¼ .69–.87, positive r’s (49) ¼ .72–.87, anxious r’s (49) ¼ .67–.69, neutral .76–.84). Correlations between positive and negative responses were significant (total ¼ .42, first ¼ .73, and endorsed ¼ .79). The SMIB subscales showed different relationships with the negative, positive, anxious, and neutral sentence completion scales. The negative subscale of the SMIB was significantly correlated with each negative measure on the sentence completion task (total ¼ .63, first ¼ .63, and endorsed ¼ .79). The positive subscale of the SMIB was not significantly correlated with any of the positive measures on the sentence completion task (total ¼ .25, first ¼ .35, and endorsed ¼ .33). The anxious subscale of the SMIB was significantly correlated with each of the anxious measures of the sentence completion task (total ¼ .46, first ¼ .54, and endorsed ¼ .47). The neutral subscale was not correlated with the neutral measures on the sentence completion task (total ¼ .13, first ¼ .15, and endorsed ¼ .01). Discussion The goal of the current study was to examine the cognitive processes involved in interpretation bias in social anxiety through a sentence completion task and a self-report measure, both using the same content. Overall, the results suggest that individuals with clinically significant social anxiety exhibit both a lack of positive bias and the presence of a negative bias at both the response generation and the response selection stages of information processing. Non-socially anxious individuals endorsed more positive and neutral self-appraisals than negative and anxious appraisals, whereas HSA individuals endorsed all appraisals equally. Between groups differences revealed that LSA individuals endorsed negative and anxious appraisals less than HSA individuals, while the groups did not differ on positive and neutral appraisals. This suggests a negative bias in the socially anxious group. When controlling for depression, only the absence of a positive bias remained an indicator of social anxiety on the sentence completion task. In contrast, the groups continued to differ in their responses on the negative and anxious scales of self-report measure even when controlling for depression. The results suggest a strong relationship among total responses, first responses, and endorsed responses on the sentence completion task. Subjects generated different types of responses within a single sentence, supporting the notion that multiple interpretations were available. In addition, on average, the second word generated by the participants was endorsed as the best-fitting response for the sentence, which raises the possibility that, contrary to expectations, the first response might not be utilized during conversations. In reallife interactions, the use of pauses, fillers (ummm or other hesitations) or empty phrases may create further time for generation of multiple interpretations followed by response selection. Thus, an important next step is to examine how these processes occur in real world-situations. In addition, it is impossible to determine from the current study whether total responses are reflective of multiple interpretations that occur prior to the final

ARTICLE IN PRESS J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

1513

resolution of ambiguity or whether the multiple responses are a later elaboration unique to the demand characteristics of the task. It is interesting to note that the relationship between self-report ratings and sentence completion responses was strong for negative and anxious scales, but low and non-significant for positive and neutral scales. Taken together with the fact that high and low-anxious individuals endorsed the positive scales at similar levels, this suggests that there are different processes influencing positive vs. negative self-appraisals. Negative self-appraisals appear to be quite related to generation and selection of responses in the face of ambiguity. These negative appraisals may be elaborations of the interpretations, which emphasize social incompetence and criticism by others, which are core features of social anxiety. In fact, negative self-appraisals or later interpretative processes may distinguish social anxiety from depression more than the negative responses generated at earlier stages of interpreting ambiguity. Consistent with the literature, elaborated negative self-appraisals are distinguishable from the negative appraisals that characterize depression (c.f., Cho & Telch, 2005). Similarly, the presence of a negative bias on self-report measures of interpretations is consistent with a number of studies (Amir, Beard, & Bower, 2005; Amir et al., 1998; Huppert et al., 2003; Stopa & Clark, 2000). The fact that depression was related to interpretation biases may provide evidence for a possible underlying cognitive mechanism that relates to the comorbidity of social anxiety and depression: the presence of a negative interpretation bias. Previous research has demonstrated a presence of negative interpretation biases in depression (Lawson & MacLeod, 1999; Rude, Valdez, Odom, & Ebrahimi, 2003), most of which has been shown with paradigms examining early stages of processing. Our data support the notion that depression is related to later stages of processing as well. Huppert et al. (2003) reported that negative interpretation bias was related to social anxiety while lack of positive bias seemed to be more related to negative affect. The current findings support the notion that later negative interpretations are related to social anxiety, but also suggest that at earlier stages, lack of positive bias relates to social anxiety. The different results for anxiety and lack of positive bias in the two studies may be due to the differences in methods of measuring bias, anxiety, or their combination. The finding that lack of positive bias in the sentence completion task is related to social anxiety is consistent with the findings of Hirsch and Mathews (2000) and with more recent data related to illusory correlation paradigms (Garner, Mogg, & Bradley, 2006). This is the first study that we are aware of which reports a unique contribution of the lack of generation of positive interpretations to social anxiety when controlling for depression. On the other hand, positive-self appraisals as measured on a self-report measure did not differ between high and low anxious groups. It is interesting to note that, in comparison with their generated and selected responses on the sentence completion task, elaborated positive self-appraisals increase in socially anxious individuals. In contrast, negative bias not only continues to be associated with social anxiety, but the distinction between groups becomes greater. Thus, ultimately, information processing in socially anxious individuals may shift from lack of positive bias (as reflected by Hirsch and Mathews) to lack of positive and presence of negative bias (as reflected by the sentence completion task), and ultimately reside with an especially negative appraisal of the self (as reflected by self-report measures). On the other hand, individuals with low levels of social anxiety appear to consider positive information throughout the stages of information processing, while they tend to diminish their emphasis on the negative. There are a number of limitations that should be mentioned. First, individuals with high levels of social anxiety were not diagnosed with social anxiety disorder, and most were not seeking treatment. Future studies may benefit from determining whether information-processing biases differ in treatment seekers compared to non-treatment seekers. Similarly, although scores of depression, anxiety, and stress were all in the normal range for low anxious individuals, they were not interviewed to determine whether they have other psychiatric disorders. Furthermore, sentence completion codes were determined by independent raters, and individuals may have coded themselves differently (i.e., being seen as quiet could have been rated positively or negatively by any given individual), and would likely have been able to report the meaning of more endings, leaving fewer responses coded as ‘‘other/uncodable’’ (Pasupuleti & Huppert, 2005). Furthermore, a number of sentences pulled for internal vs. external responses (e.g., ‘‘someone is looking at XXX’’). Those sentences ended up being less likely to discriminate high and low individuals and elicited uncodeable responses. It would be useful to examine the same material in a lexical or grammatical decision task similar to that used by Hirsch and

ARTICLE IN PRESS 1514

J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

Mathews (1997) in order to examine earlier stages of information processing. Furthermore, it is possible that the self-report measure did not reflect the lack of positive bias due to a ‘‘ceiling effect’’ in that modesty or social desirability may have prevented low anxious individuals from endorsing an overly positive picture of themselves, or allowed high anxious individuals to put themselves in a somewhat positive light. We are currently examining whether all three processes can be reflected within a single subject by including reaction time paradigms in addition to sentence completion and self-report measures. At the same time, more information is needed to determine the differences in processes utilized for self-report compared with other measures of interpretation biases. For example, the ways in which they differ include the time and opportunity available for slower reflective processes to influence responses, as well as in their transparency and resultant influence of prior beliefs about the self or self-presentation motivation. This study builds on previous studies, which have examined interpretation bias in social anxiety. With the accumulating data regarding the absence of positive interpretation bias and presence of negative interpretation bias, there are a number of future directions which will continue to elucidate the processes involved in the maintenance of social anxiety and provide future directions for treatments and research. For example, how do reaction time or other measures of interpretation relate to later stages of processing and to one another? Are these biases differentially related to positive or negative affect? How does stress impact these biases? Does it impact more negative or positive interpretation? If both lack of positive and presence of negative interpretation biases are present, should treatment be focused first on one and then on the other, or is focus on the earlier one sufficient? The last question has both clinical (c.f., Alden, Mellings, & Laposa, 2004), and theoretical implications related to the debate regarding suppression vs. response competition (c.f., Gernsbacher & St. John, 2001; MacLeod et al., 2003). Acknowledgements The authors would like to thank Richard G. Heimberg for comments on a previous draft, and Rustin Simpson for collecting and coding data. This paper was funded by NIMH grant 5K23MH064491 awarded to the first author.

References Alden, L. E., Mellings, T. M. B., & Laposa, J. M. (2004). Framing social information and generalized social phobia. Behaviour Research and Therapy, 42(5), 585–600. Amir, N., Beard, C., & Bower, E. (2005). Interpretation bias and social anxiety. Cognitive Therapy and Research, 29(4), 433–443. Amir, N., Beard, C., & Przeworski, A. (2005). Resolving ambiguity: The effect of experience on interpretation of ambiguous events in generalized social phobia. Journal of Abnormal Psychology, 114(3), 402–408. Amir, N., Foa, E. B., & Coles, M. E. (1998). Negative interpretation bias in social phobia. Behaviour Research and Therapy, 36(10), 945–957. Antony, M. M., Bieling, P. J., Cox, B. J., Enns, M. W., & Swinson, R. P. (1998). Psychometric properties of the 42-item and 21-item versions of the depression anxiety stress scales in clinical groups and a community sample. Psychological Assessment, 10(2), 176–181. Baker, S. L., Heinrichs, N., Kim, H., & Hofmann, S. G. (2002). The Liebowitz social anxiety scale as a self-report instrument: A preliminary psychometric analysis. Behaviour Research and Therapy, 40(6), 701–715. Brown, T. A., Chorpita, B. F., Korotitsch, W., & Barlow, D. H. (1997). Psychometric properties of the depression anxiety stress scales (DASS) in clinical samples. Behaviour Research and Therapy, 35(1), 79–89. Cho, Y., & Telch, M. J. (2005). Testing the cognitive content-specificity hypothesis of social anxiety and depression: An application of structural equation modeling. Cognitive Therapy and Research, 29(4), 399–416. Clark, D. M. (2001). A cognitive perspective on social phobia. In R. W. Crozier, & L. E. Alden (Eds.), International handbook of social anxiety: Concepts, research and interventions relating to the self and shyness (pp. 405–430). New York: Wiley. Conner, K. M., Davidson, J. R. T., Churchill, L. E., Sherwood, A., Foa, E., & Weisler, R. H. (2000). Psychometric properties of the Social Phobia Inventory (SPIN): New self-rating scale. British Journal of Psychiatry, 176, 379–386. Crawford, J. R., & Henry, J. D. (2003). The depression anxiety stress scales (DASS): Normative data and latent structure in a large nonclinical sample. British Journal of Clinical Psychology, 42(2), 111–131. Fischer, D. G., & Fick, C. (1993). Measuring social desirability: Short forms of the Marlowe-Crowne social desirability scale. Educational and Psychological Measures, 57, 424–427. Franklin, M. E., Huppert, J. D., Langner, R., Leiberg, S., & Foa, E. B. (2005). Interpretation bias: A comparison of treated social phobics, untreated social phobics, and controls. Cognitive Therapy and Research, 29, 289–300.

ARTICLE IN PRESS J.D. Huppert et al. / Behaviour Research and Therapy 45 (2007) 1505–1515

1515

Fresco, D. M., Coles, M. E., Heimberg, R. G., Leibowitz, M. R., Hami, S., Stein, M. B., et al. (2001). The Liebowitz social anxiety scale: A comparison of the psychometric properties of self-report and clinician-administered formats. Psychological Medicine, 31(6), 1025–1035. Garner, M., Mogg, K., & Bradley, B. P. (2006). Fear-relevant selective associations and social anxiety: Absence of a positive bias. Behaviour Research and Therapy, 44, 201–217. Gernsbacher, M. A. (1991). Cognitive processes and mechanisms in language comprehension: The structure building framework. In G. H. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory, Vol. 27 (pp. 217–263). San Diego: Academic Press. Gernsbacher, M. A., & St. John, M. F. (2001). Modeling suppression in lexical access. In D. S. Gorfein (Ed.), On the consequences of meaning selection: Perspectives on resolving lexical ambiguity. Decade of behavior (pp. 47–65). Washington, DC: American Psychological Association. Gilboa-Schechtman, E., Presburger, G., Marom, S., & Hermesh, H. (2005). The effects of social anxiety and depression on the evaluation of facial crowds. Behaviour Research and Therapy, 43, 467–474. Gotlib, I. H., Kasch, K. L., Traill, S., Joorman, J., Arnow, B. A., & Johnson, S. L. (2004). Coherence and specificity of informationprocessing biases in depression and social phobia. Journal of Abnormal Psychology, 113(3), 386–398. Hirsch, C., & Mathews, A. (1997). Interpretive inferences when reading about emotional events. Behaviour Research and Therapy, 35(12), 1123–1132. Hirsch, C. R., & Mathews, A. (2000). Impaired positive inferential bias in social phobia. Journal of Abnormal Psychology, 109(4), 705–712. Huppert, J. D., & Foa, E. B. (2004). Maintenance mechanisms in social anxiety: An integration of cognitive biases and emotional processing theory. In J. Yiend (Ed.), Cognition, emotion and psychopathology: Theoretical, empirical and clinical directions (pp. 213–231). New York: Cambridge University Press. Huppert, J. D., Foa, E. B., Furr, J. M., Filip, J. C., & Mathews, A. (2003). Interpretation bias in social anxiety: A dimensional perspective. Cognitive Therapy and Research, 27(5), 569–577. Kutas, M., & Hillyard, S. A. (1984). Brain potentials during reading reflect word expectancy and semantic association. Nature, 307, 161–163. Lawson, C., & MacLeod, C. (1999). Depression and the interpretation of ambiguity. Behaviour Research and Therapy, 37(5), 463–474. Liebowitz, M. R. (1987). Social phobia. Modern Problems in Pharmacopsychiatry, 22, 141–173. Lovibond, P. F., & Lovibond, S. H. (1995). The structure of negative emotional states: Comparison of the depression anxiety stress scales (DASS) with the beck depression and anxiety inventories. Behaviour Research and Therapy, 33(3), 335–343. MacLeod, C. M., Dodd, M. D., Sheard, E. D., Wilson, D. E., & Bibi, U. (2003). In opposition to inhibition. The Psychology of Learning and Motivation, 43, 163–214. Mattick, R. P., & Clarke, C. J. (1998). Development and validation of measures of social phobia scrutiny fear and social interaction anxiety. Behaviour Research and Therapy, 36(4), 455–470. McNamara, D. S., & McDaniel, M. A. (2004). Suppressing irrelevant information: Knowledge activation or inhibition? Journal of Experimental Psychology: Learning, Memory, & Cognition, 30(2), 465–482. Mennin, D. S., Fresco, D. M., Heimberg, R. G., Schneier, F. R., Davies, S. O., & Liebowitz, M. R. (2002). Screening for social anxiety disorder in the clinical setting: Using the Liebowitz social anxiety scale. Journal of Anxiety Disorders, 16(6), 661–673. Pasupuleti, R. V., Huppert, J. D. (November, 2005). Subject vs. independent raters: Can one evaluate one’s own mind? Poster presented at the 39th annual meeting of the Association for Behavioral and Cognitive Therapies, Washington, DC. Rapee, R. M., & Heimberg, R. G. (1997). A cognitive-behavioral model of anxiety in social phobia. Behaviour Research and Therapy, 35(8), 741–756. Ries, B. J., McNeil, D. W., Boone, M. L., Turk, C. L., Carter, L. E., & Heimberg, R. G. (1998). Assessment of contemporary social phobia verbal report instruments. Behaviour Research and Therapy, 36(10), 983–994. Rude, S. S., Valdez, C. R., Odom, S., & Ebrahimi, A. (2003). Negative cognitive biases predict subsequent depression. Cognitive Therapy and Research, 27(4), 415–429. Safren, S. A., Turk, C. L., & Heimberg, R. G. (1998). Factor structure of the social interaction anxiety scale and the social phobia scale. Behaviour Research and Therapy, 36(4), 443–453. Stopa, L., & Clark, D. M. (2000). Social phobia and interpretation of social events. Behaviour Research and Therapy, 38(3), 273–283. Strahan, R., & Gerbasi, K. C. (1972). Short, homogeneous versions of the Marlowe-Crowne social desirability scale. Journal of Clinical Psychology, 28(2), 191–193. Watson, D., Clark, L. A., Weber, K., Assenheimer, J. S., Strauss, M. E., & McCormick, R. A. (1995). Testing a tripartite model: II. Exploring the symptom structure of anxiety and depression in student adult and patient samples. Journal of Abnormal Psychology, 104(1), 15–25. Williams, J. N., & Colombo, L. (1995). Constraints on the range of context-independent priming from ambiguous words. Psychological Research, 58(1), 38–50.